Fairness-aware News Recommendation with Decomposed Adversarial Learning

News recommendation is important for online news services. Most news recommendation methods model users' interests from their news click behaviors. Usually the behaviors of users with the same sensitive attributes have similar patterns, and existing news recommendation models can inherit these biases and encode them into news ranking results. Thus, their recommendation results may be heavily influenced by the biases related to sensitive user attributes, which is unfair since users cannot receive sufficient news information that they are interested in. In this paper, we propose a fairness-aware news recommendation approach with decomposed adversarial learning and orthogonality regularization, which can alleviate unfairness in news recommendation brought by the biases of sensitive user attributes. For model training, we propose to learn a bias-aware user embedding that captures the bias information on user attributes from click behaviors, and learn a bias-free user embedding that only encodes attribute-independent user interest information for fairness-aware news recommendation. In addition, we propose to apply an attribute prediction task to the bias-aware user embedding to enhance its ability on bias modeling, and we apply adversarial learning to the bias-free user embedding to remove the bias information from it. Moreover, we propose an orthogonality regularization method to encourage the bias-free user embeddings to be orthogonal to the bias-aware one to further purify the bias-free user embedding. For fairness-aware news ranking, we only use the bias-free user embedding. Extensive experiments on benchmark dataset show that our approach can effectively improve fairness in news recommendation with acceptable performance loss.